In a week chock-a-block with copyright news, a popular pastime in Washington, D.C., has been trying to figure out whether the Copyright Office’s (pre-publication) report on fair use and AI training, which was released late Friday night, had been a factor in the Trump administration’s unceremonious (attempted) firing of the Register of Copyrights, Shira Perlmutter, later that weekend. I won’t enter that fray (though I’m happy to speculate over a beer, if you’re buying). But I do want to answer an important related question: what was the real conclusion of the Copyright Office’s report, anyway? 

Despite all the noise about the report’s muddled fair use reasoning, the real bottom line is quite simple—Congress does not need to pass new laws to accommodate AI in the copyright system. This is stated clearly in the conclusion of the report, with respect to both fair use and licensing:

While the use of copyrighted works to power current generative AI systems may be unprecedented in scope and scale, the existing legal framework can address it as in prior technological revolutions. The fair use doctrine in particular has served to flexibly accommodate such change. We believe it can do so here as well.…

Given the robust growth of voluntary licensing, as well as the lack of stakeholder support for any statutory change, the Office believes government intervention would be premature at this time.

It may disappoint the plaintiffs in pending cases, but the Office’s statutory responsibility in terms of providing legal opinions is to advise Congress, the branch of government in which it sits, not to file untimely, over-long, unsolicited amicus briefs in district courts. As I argue at more length in my analysis of the report, courts aren’t likely to be persuaded, if they consider it at all. Congress is the branch that asked for guidance on these issues, however, and they should heed the Office’s advice.

That means abandoning bills like the TRAIN and COPIED Acts, which preempt fair use and upset the balance of the copyright framework. Related bills like NO FAKES also merit a second look to the extent they create new IP rights that could interfere with the established copyright system. State legislators should also take heed. Bills like California’s AB 412 intrude into the copyright system on the assumption that all AI training is infringement—an assumption the Office’s report rejects. 

In the rush to suss out whether the Office sided with AI companies or content companies on the fair use question, many observers missed the real point: the Office told Congress to leave these issues to the courts. Here’s hoping they listen.